Get a weekly rundown of the latest AI models and research... subscribe! https://aimodels.substack.com/

Supir

cjwbw

AI model preview image
The Supir model is an image-to-image model used for image restoration. It practices Model Scaling to enhance photo-realistic images in the wild. In this version, it employs LLaVA-13b for image captioning. The model takes in an image URL, various configuration settings controlling the restoration process, a narrative ("a_prompt") that directs the restoration toward a higher quality, detailed outcome, and another narrative ("n_prompt") that refrains the model from producing certain undesirable effects. After processing, the model outputs an enhanced image URL. In other words, the model can enhance and restore details to photos, correct color, improve sharpness, and remove undesirable elements such as blurs, jpeg artifacts, and deformations. It can also adjust image features to match specific descriptions, like an image taken with a specific camera or an image with ultra-HD quality.

Use cases

The SUPIR AI model is designed to practice model scaling and offers capabilities for photo-realistic image restoration in diverse real-world scenarios. Given its capacity to handle high-definition image manipulation, the model can be employed to produce hyper-detailed photographs with intricate levels of detailing such as skin pore detailing or color grading. Potential applications could include professional photography, digital artistry, cinema, or ad agencies where there is a high premium on delivering ultra HD, detail-oriented imagery. The model can also be utilized to transform lower-quality images by removing deformities, blurring, or artifacts, thus providing a solution for enhancing images for personal use, social media posts, or historical photo restoration. Furthermore, its use of a wavelet color fix type may enable color correction services in photography or film grading. More so, its capacity to convert photographs into different art styles like oil painting, cartoon, or 3D render could serve artists, illustrators, or the graphic design industry. Lastly, the image-to-image feature could potentially be integrated into software or apps as a tool for users to manipulate and enhance their digital imagery.

Image-to-Image

Pricing

Cost per run
$-
USD
Avg run time
-
Seconds
Hardware
-
Prediction

Creator Models

ModelCostRuns
Eimis_​anime_​diffusion$?0
Dreambooth Pikachu$0.08195513
Cutie$?171
Night Enhancement$0.0104538,658
Controlvideo$?1,834

Similar Models

Try it!

You can use this area to play around with demo applications that incorporate the Supir model. These demos are maintained and hosted externally by third-party creators. If you see an error, message me on Twitter.

Currently, there are no demos available for this model.

Overview

Summary of this model and related resources.

PropertyValue
Creatorcjwbw
Model NameSupir
Description

Practicing Model Scaling for Photo-Realistic Image Restoration In the Wild....

Read more Β»
TagsImage-to-Image
Model LinkView on Replicate
API SpecView on Replicate
Github LinkView on Github
Paper LinkView on Arxiv

Popularity

How popular is this model, by number of runs? How popular is the creator, by the sum of all their runs?

PropertyValue
Runs25,410
Model Rank
Creator Rank

Cost

How much does it cost to run this model? How long, on average, does it take to complete a run?

PropertyValue
Cost per Run$-
Prediction Hardware-
Average Completion Time-